Deutsches Institut für Wirtschaftsforschung

www.diw.de

Marc Vothknecht • Sudarno Sumarto

Berlin, February 2011

Beyond the Overall Economic Downturn: Evidence on Sector-specific Effects of Violent Conflict from Indonesia

1105

Discussion Papers

Opinions expressed in this paper are those of the author(s) and do not necessarily reflect views of the institute. IMPRESSUM © DIW Berlin, 2011 DIW Berlin German Institute for Economic Research Mohrenstr. 58 10117 Berlin Tel. +49 (30) 897 89-0 Fax +49 (30) 897 89-200 http://www.diw.de ISSN print edition 1433-0210 ISSN electronic edition 1619-4535 Papers can be downloaded free of charge from the DIW Berlin website: http://www.diw.de/discussionpapers Discussion Papers of DIW Berlin are indexed in RePEc and SSRN: http://ideas.repec.org/s/diw/diwwpp.html http://www.ssrn.com/link/DIW-Berlin-German-Inst-Econ-Res.html

Beyond the Overall Economic Downturn: Evidence on Sector-specific Effects of Violent Conflict

from Indonesia

Marc Vothknecht* German Institute for Economic Research (DIW Berlin) and

Humboldt University Berlin

Sudarno Sumarto

SMERU Research Institute, Jakarta

Abstract

This paper analyses the impact of violent conflict on economic growth using micro-level data from Indonesia. We compile a panel dataset at district level for the period 2002-2008, and disentangle the overall negative economic effect of violent conflict into its sectoral components. Our results reveal substantial differences across sectors, with the most detrimental impact evident in manufacturing industries and the service sectors. Further, the short-run impacts on growth appear to be only temporal, and some evidence for the ‘phoenix effect’ in the early post-conflict period is found. The construction sector, in particular, recovers soon once conflict ends, while manufacturing industries and the finance sector appear especially reliant on a lasting peace. A series of alternative specifications confirm the main findings of the analysis.

Keywords: Violent Conflict, Economic Growth, Indonesia

JEL Codes: O11, F51

* Corresponding author: mvothknecht@diw.de. We would like to thank Tilman Brück, Joppe de Ree and Günther Schulze as well as seminar participants in Berlin, Bonn, Glasgow and Washington DC for their comments on earlier versions of this paper. All remaining errors are ours.

INTRODUCTION

Indonesia, the largest and most populous Southeast Asian nation, emerged with a new

democratic government following the 1998 fall of President Suharto. The collapse of

Suharto’s long and authoritarian New Order regime was followed by a wave of high-profile

violence, with separatist conflicts in Aceh and East Timor, religious conflicts in Maluku and

Sulawesi, and ethnic conflicts in Kalimantan and Nusa Tenggara Timur. One decade after the

end of the New Order, some progress in restoring peace and political stability has been

achieved and Indonesia is moving toward a more stable democratic government while the

number of headline-making conflicts has sharply decreased.

This paper employs exceptionally detailed data sources to analyze the potentially varying

effects of violent conflict on different areas of the Indonesian economy. While existing

empirical evidence on the conflict-growth nexus mostly comes from cross-country studies and

is based on aggregate GDP data, our approach allows going beyond the (well-known) overall

negative impact of war. For this purpose, we combine district-level data on GDP composition

with nationwide information on conflict occurrence in 2002, 2005, and 2008.

The panel analysis provides important insights into sector-specific vulnerabilities during

violent conflict and in the early recovery phase. We find, first, that activities depending on

capital and transactions are particularly affected by violent conflict. Second, conflict-related

declines in economic growth are of temporary nature, with some areas of the economy

recovering faster than others once conflict ends. Third, the results confirm that the scale of the

economic downturn depends on the intensity and type of violence, rather than the mere

existence of mostly non-violent or low intensity conflicts. Finally, spillover effects from

violence in neighboring regions are less clear and appear to only affect transaction-intensive

industries, most notably transport industries and financial services.

The paper is structured as follows. The next section reviews the existing literature on growth

and violence, discusses potential transmission channels from violence to growth, and derives

some hypotheses on the sector-specific impacts of violent conflict. We then provide

background information to the conflicts under consideration, present the data, and discuss our

estimation strategy. This is followed by some descriptive statistics and the presentation of the

results from the regression analysis.

2

BACKGROUND

Growth Empirics

Explaining differences in economic growth and, more generally, in income levels across

countries and over time is one of the most important, although one of the more difficult, tasks

in economic research. The existing literature is vast and constantly amended by new

contributions to both theory and empirics. Solow’s 1956 model is still the major theoretical

reference and basic workhorse of growth empirics. Based on a neoclassical production

function, long-run per capita incomes in this model depend on savings and population growth

rates as well as technological progress. Countries with similar characteristics in this respect

are expected to converge to a common steady state, with only transitory income differences

between these countries.

This neoclassical theory of convergence is challenged by real world divergent growth

experiences. The permanent income differences between countries are attributed to structural

and institutional heterogeneities with empirical work finding a wide range of economic,

political, social and geographical factors that impact economic growth.1 The validity of this

large amount of work, however, with a variety of different and sometimes opposing results, is

increasingly called into question. Inter alia, criticism includes arbitrary model specifications

without theoretical underpinning, issues of data comparability, inappropriate econometric

methodologies, limitations of cross-country analyses, and insufficient treatment of

endogenous explanatory variables (see e.g. Bosworth and Collins, 2003). Some work tries to

account for these shortcomings, with Durlauf et al. (2005) identifying a set of explanatory

variables consistently found to be related to economic growth. While the focus of this study is

on violent conflict rather than on the growth process as such, cross-country evidence on these

factors is briefly reviewed.

Today, scholars widely agree on the importance of institutional settings for economic

development. Acemoglu et al. (2004) stress the role of economic institutions, such as well-

defined property rights and the presence of functioning markets for economic outcomes. The

quality of both economic and political institutions, as well as the impact these have on

economic growth, is assessed in a wide range of studies. Acemoglu et al. (2001), for instance,

tackle concerns of endogeneity by using the mortality of colonial settlers as an instrument for

1 Durlauf et al. (2005) report a total of 145 potential determinants of economic growth that are studied in the literature.

3

today’s institutional quality and confirm a decisive impact on growth.

Besides historical legacies, ethnic diversity is often seen as another explanation for cross-

country differences in the effectiveness of public policies. Heterogeneous preferences across

ethnic groups are likely to impede agreement on policy decisions and result in lower spending

on public goods (Alesina et al. 1999). Sound government policy thereby plays a particularly

decisive role in the area of infrastructure. Efficient investment, for example in

telecommunications, transport, water or energy systems, facilitates economic activity and is

found to contribute substantially to GDP growth (Esfahani and Ramírez, 2003).

The relationship between education and economic growth is more contested. While a positive

link between individual education and income is confirmed by a large body of microeconomic

studies (see e.g. Psacharapoulos and Patrinos, 2004), cross-country evidence is mixed. Recent

macro-level analyses on the impact of educational attainment on economic growth fail to

establish a significant relationship (see Easterly and Levine (2001) for a discussion). Potential

explanations include differences between social and private returns to education,

measurement errors and limited data comparability.

Other country characteristics that are identified as fundamental growth determinants, such as

climatic conditions (Masters and McMillan, 2001), geographic isolation (Frankel and Romer,

1999), or culture and religion (Barro and McCleary, 2003), are rather time-invariant and can

barely be influenced, at least in the short term. In general, cross-country studies tend to focus

on the factors that drive long term growth and rarely provide for short term interventions. A

distinction between short term growth dynamics and long term equilibrium effects could

therefore add valuable insights for policy makers (Rao and Corray, 2009).

Evidence from cross-country studies often also suffers from restrictive assumptions. Pooling

across countries implicitly assumes a universal growth process and homogeneous growth

parameters, i.e. identical underlying production functions and technological improvements.

With most samples, however, consisting of very heterogeneous countries at different stages of

development and with diverse economic structures, cross-country parameter estimates, at best,

display inter-country averages, but tell little about the evolution of growth for individual

countries (Bos et al., 2010).

Therefore, case studies using country specific time series or panel data are more appropriate

for analyzing country specific growth experiences. Common institutional and legal

frameworks plus the consistency of data sources within individual countries contribute to a

higher comparability of growth processes. Disaggregated GDP data often enables

4

investigation into inter-industry differences. Empirical research in this field has focused, inter

alia, on the determinants of growth rates in specific sectors, including, among others, Sapio

and Thoma (2006) on US manufacturing, and Bosma, Stam and Schutjens (2010) on Dutch

manufacturing and services; as well as the contribution of specific sectors to overall economic

development, including Balaguer and Cantavella-Jordá (2002) on Spanish tourism.

Evidence from developing countries is relatively scarce. For Indonesia, Garcia and

Soelistianingsih (1998) analyze the evolution of provincial incomes under the New Order

regime and find that poorer provinces catch up, especially through investments in human

capital. Using district-level data for the period 1993-2005, McCulloch and Sjahrir (2008)

confirm the hypothesis of relative convergence and the positive growth impact of a better

educated labor force. None of these studies accounts for the effects of violent conflict on the

local economy.

Impact of Violent Conflict on Growth

Violent conflicts affect economic outcomes mainly through the destruction of human and

physical capital, shifts in public spending and private investment, as well as the disruption of

economic activities and social life (see Blattman and Miguel (2010) for a summary). The

specific impacts depend on each conflict’s singular characteristics: it is not just the type of

conflict, but also its intensity, duration, and geographical spread that shapes its economic

consequences. We also expect that violent conflicts affect individual economic sectors

differently, given differing characteristics.

Quantitative evidence on the overall growth effect of conflicts mostly comes from cross-

country datasets. Collier’s (1999) analysis establishes a substantially negative link between

civil war duration and economic growth. His approach, however, is criticized for not

considering variations in the scale and scope of conflict. Imai and Weinstein (2000) stress the

importance of the geographical spread of conflicts, and conclude that wide-spread civil wars

affect economic growth five times more than conflicts fought in small parts of the country.

Bodea and Elbadawi (2008) distinguish different levels of political violence and find

particularly negative growth impacts for civil wars, relative to the less severe effects resulting

from riots or coups. Consistent with this, Koubi (2005) finds that the impact that civil war has

on long-run economic growth to be proportional to conflict severity in terms of conflict-

related human losses.

5

Single country evidence on the growth impact of violence is relatively scarce. Abadie and

Gardeazabal (2003) show a negative impact of the ETA terrorist conflict on economic growth

in the Basque Country. Dependent on the intensity of terrorism, Basque per capita GDP

declined, on average, by 10 percentage points relative to other Spanish regions. Lopez and

Wodon (2005) use time series data to estimate the impact of the 1994 genocide on the

evolution of Rwanda’s per capita GDP. Based on outlier detection and correction, their results

indicate a significant loss in GDP in the short term, though no impact is found on post-

genocide growth rates. Arunatilake et al. (2001) assess the direct and indirect costs of the Sri

Lankan civil war between 1983 and 1996, with their estimates of total costs adding up to

twice the country’s 1996 GDP.

Existing literature on Indonesian conflicts mostly deals with the socioeconomic determinants

of violent conflict and not with its consequences (Tadjoedin and Murshed, 2007). Barron,

Kaiser and Pradhan (2009) use the 2003 PODES village dataset to analyze correlations of

local violent conflicts with several socioeconomic variables, including unemployment,

inequality, and natural disasters. Anecdotal evidence reveals that the economic potential of

conflict prone areas is not fully realized due to conflicts. Most production in these areas is

disrupted and those who trying to export products face barriers such as illegal taxes or fees

imposed by civil servants and military personnel (Mawardi, 2006).

Often, the economic impact of violent conflicts is not limited to the conflict area as the impact

spills over to neighboring regions and countries. Refugee flows increase labor supply and

reduce per capita incomes, trade flows are disrupted, government spending is diverted to non-

productive security measures, and foreign direct investments might be shifted away from

entire regions that are perceived as insecure. Murdoch and Sandler’s (2002) analysis of spatial

dispersion finds significant negative growth impacts of civil wars for contiguous neighbors,

especially in the short term. Using different data and methodology, De Groot (2010) confirms

the negative economic effects of violent conflict for direct neighbors, but finds positive

spillovers for non-contiguous areas.

The overall negative effects of violent conflict on economic development are hence well

documented and seem shaped by the conflict’s type, duration, and geographical spread.

However, our understanding of, and especially empirical evidence on, the transmission

mechanisms from conflict to growth remain limited. In what follows, we characterize the

drivers of the economic downturn in conflict-ridden societies and present the existing

evidence on the varying effects of conflict across economic activities.

6

Transmission Channels from Conflict to Growth

First and foremost, the level and growth rate of the capital stock are negatively affected by

conflict-related damage and reduced investment. The destruction or dislocation of production

facilities and key infrastructure, most notably in the areas of transportation, communication

and energy, impedes economic activity. Private agents become involved in dissaving and

portfolio substitution as perceived risk increases and as investments in regions not affected by

violent conflict offer higher relative returns with lower risk. Government spending on

productive activities is likely to decrease due to an eroding tax base and hampered tax

administration on the one hand, and the diversion of public funds to military and security

expenditures on the other. Poor macroeconomic policy, with rising budget deficits and

increasing inflation, tends to further hamper economic growth.

Empirical work mostly focuses on changes in public and private investment. Svensson (1998),

for instance, identifies a potential channel from political instability and associated weak

property rights to reduced private investment. A negative impact of political violence on

private savings and domestic investments is also found by Fielding (2000, 2004) in his

analyses of the macroeconomic impacts of the Israeli Intifada. Gupta et al. (2004) reveal

adverse effects of armed conflict on tax revenues and public investments: higher government

spending on defense is associated with macroeconomic instability and a diversion of

resources away from socially and economically productive sectors. Knight et al. (1996) relate

this growth-retarding impact of rising military expenditures to negative effects on capital

formation and resource allocation. Finally, Imai and Weinstein (2000) conclude that

reductions in private investments affect economic growth much more than downturns in

public investment.

Assessing the growth impact of other conflict phenomena, such as the loss of human capital

or changes in economic activities, is more difficult. People living in conflict zones face

increased risks of being injured or killed. The better off (and better educated) are more likely

to flee to neighboring regions or countries. In the longer term, human capital accumulation is

affected by the destruction of educational infrastructure, the absence of teachers, and lower

government spending on education during times of war. The negative long term consequences

of civil wars for aggregate health and education measures are documented by an increasing

7

number of studies.2

In the short term, conflict-related deaths and injuries, emigration, displacement, and forced

conscription reduce the available labor force and are likely to affect productivity. However,

systematic empirical evidence on the impact of violent conflict on labor markets hardly exists.

One exception is the assessment by Arunatilake et al. (2001) of the costs of the civil war in

Sri Lanka. Their estimates of the costs related to displacement and lost human capital sum up

to around five percent of the estimated total costs of war.

More is known about conflict-related distortions of economic behavior, i.e. impacts on

economic activity and market exchange. Heightened insecurity, threatened property rights,

and the suppression of civil liberties result in an unsafe business environment. Transaction

costs increase, trust deteriorates, economic activities are disrupted and market exchange is

limited in war-torn societies. Micro-level studies document people’s retreat into less

vulnerable economic activities in conflict regions. Deininger (2003), for example, shows that

civil strife in Uganda during the 1990s reduced off-farm investments and led to a shift toward

subsistence farming with less market integration. Similarly, Verpoorten (2009) finds that the

Rwandan households most affected by violence were prevented from selling their cattle, the

usual response to adverse income shocks, due to unsafe roads and the related inaccessibility of

markets.

Further, violent conflict hampers not only domestic exchange, but international trade flows as

well. Blomberg and Hess (2006) find that global trade flows are impeded to a greater extent

by violent conflicts than by traditional tariff barriers. According to their estimates, bilateral

trade flows for conflict-affected countries decline by up to 40 percent. The strongly negative

impact of war on international trade is confirmed by Glick and Taylor’s (2005) analysis

covering 1870-1997.

Sector-specific Evidence and Expectations

Given the substantial and diverse effects of violent conflict, we expect varying degrees of

vulnerability across different economic sectors. Economic activities that are strategically

2 Ghobarah et al. (2003) examine the impact of civil wars on public health; Lai and Thyne’s (2007) cross-country study finds devastating war effects on educational systems. Growing evidence on the human capital costs of violent conflict is also available at the micro-level; see Households in Conflict Network, http://www.hicn.org, for recent studies.

8

important to the belligerent parties might also become direct targets, such as relevant

production facilities or the transportation and logistics sectors. Empirical evidence on these

sector-specific effects of violent conflict is generally rare and comes mostly from single-

sector case studies, as disaggregated growth measures from war-torn economies are often not

available. Collier (1999) argues that those activities, which supply or are intensive in either

capital or transactions, are particularly vulnerable. Sectors less dependent on capital and

transactions should be relatively less affected, which applies to and is confirmed by the

observed rise in subsistence agricultural activities in times of conflict.

The overall impact of conflict on primary sector growth, however, seems ambiguous.

Destroyed assets, landmine contamination, as well as a shortage of labor and capital are likely

to impede agricultural development. The production of (cash) crops and livestock requires

future investments and is expected to decline due to shortened planning horizons for farmers.

Dramatic losses of livestock are documented for several civil wars.3 Teodosijević (2003)

compares growth rates of agricultural production in 38 conflict-affected countries and finds

substantially lower outcomes during violent conflict in comparison with the pre-war period.

Generally, concerns of endogenous economic outcomes with respect to violent conflict

particularly apply to the primary sector. Being central to many societies in the developing

world, environmental scarcities and food insecurities are among the main triggers of violent

conflicts. Prospects for agricultural development might therefore be structurally lower in

conflict-prone regions.

Similarly, the link between natural resource wealth and violent conflict is likely to run both

ways. Scholars find that the production of oil and ‘lootable resources’, such as gemstones and

drugs, to be associated with conflict incidence, specifically separatist conflicts (see Ross

(2004) for a summary). Violent conflicts, in turn, tend to increase economic dependency on

natural resources: while other sectors of the economy collapse, natural resources are immobile

and represent an often important source of revenue for governments and/or rebels. Official

growth rates of the mineral sector then depend on the legal status of the profiteer and the

perceived legitimacy of extraction.

Turning to the industrial sector and applying Collier’s (1999) concept of war vulnerability, we

expect manufacturing and construction industries to be particularly affected. Manufacturing is

3 Brück (1997) estimates a loss of 80% of the cattle stock in the Mozambican civil war. Similar devastation is found for Uganda (Annan, Blattman and Horton, 2006) and Rwanda (FAO, 1997).

9

often intensive in both physical and human capital, requires complex coordination with

contractors and vendors, and is therefore especially reliant on the institutional environment. In

his analysis of conflict-ridden developing countries, Depetris Chauvin and Rohner (2009)

points to negative effects of war for exporting industries due to conflict-related trade barriers.

The collapse of the domestic demand for investment goods, both from state and private

agents, would also, and in particular, affect the construction sector.

The service sector, finally, encompasses a broad range of economic activities, ranging from

traditional services – house cleaning, food services, barber shops, and the like – to trade,

transport, tourism, the social sector, as well as finance and business services. Finance is

expected to suffer from capital flight and falling demand for transactions, with the latter also

applying to trade and transport industries. Likewise, tourism is particularly sensitive to violent

events: Neumayer’s (2004) cross-country analysis reveals strong negative impacts of political

violence on tourist arrivals, with even stronger impacts in the long term compared to

contemporaneous effects.

Increasing military expenditures during violent conflict involve shifts in government spending

often to the detriment of social services. Traditional services tend to be location-specific and

are less dependent on physical capital and long term investments. Assumed to be relatively

invulnerable to war, such local services might nevertheless suffer from the overall economic

decline. Generally, the lack of employment opportunities in conflict times often leads to an

increase in informal businesses. These activities, such as petty trade, peddling work, or the

selling of food are difficult to capture in official statistics.

One important issue omitted so far is the dynamic aspect of violent conflict, i.e. varying

economic impacts at different stages of conflict. Disregarding conflict intensity, the negative

shock related to the outbreak of violence might have a more detrimental growth impact than

an ongoing conflict. Cerra and Saxena (2008), for instance, show that the initial economic

decline in countries affected by civil war is substantial, but they expect at least partial

recovery of output levels after a relatively short period of time. Contrary to the convergence

predictions of neoclassical growth models, however, a complete catch-up to pre-war levels in

the post-conflict period is not observed.

Building on the Phoenix factor theory (Organski and Kugler, 1977), which refers to the post-

conflict restoration of economic activities, we still expect positive prospects for economic

growth in the recovery process. Efforts of reconstruction, for instance, might spark a

construction boom once violence and insecurity decline (Collier, 2009). Recovery of other

10

areas of the economy, such as the tourism industry or the financial sector, might be more

reliant on the rebuilding of trust and a peaceful environment, which takes time.

CONFLICT AND DATA DESCRIPTION

Background to the Violence

Before turning to the data used to analyze the sector-specific effects of violent conflict in the

early years of the post-Suharto era, we briefly characterize the most severe conflicts of that

era. Most international attention is paid to the conflict between the Free Aceh Movement

(GAM) rebels in Aceh and the Indonesian government (see Reid (2006) for an in-depth

analysis of the conflict). This separatism fight began with GAM’s declaration of an

independent Aceh in 1976 and lasted for almost 30 years, with an estimated 12,000 killed

during the conflict. As a province rich in oil and gas, with a long history of independence and

a regional specific Muslim character, Acehnese aspirations for political autonomy were based

on economic as well as historical and religious motivation. The New Order regime under

Suharto responded to the secessionist movement with increasingly repressive measures, while

at the same time taking a lion’s share of the province’s significant revenues from oil and gas.

The fall of Suharto in 1998 raised hope for an end to the conflict and peace talks between the

democratically elected Indonesian Government and GAM were initiated with special

autonomy status offered to the province. Concurrently, fighting between GAM rebels and the

Indonesian military continued, and even intensified between 1999 and 2002, resulting in a

large death toll, especially among civilians. In May 2003, when peace talks finally failed, the

Indonesian government imposed martial law in Aceh and started a major military offensive to

ultimately weaken the rebel movement.

Growing battle fatigue among GAM members combined with the newly elected President

Yudhoyono’s and Vice President Kalla’s political will to end the conflict contributed to a

tentative resumption of the peace process in late 2004. It was in this context that Aceh was hit

by the devastating tsunami in late December 2004. It struck much of the western and northern

coast of the province. Both belligerent sides agreed to cease hostilities in order to facilitate the

recovery process, and the subsequent peace talks resulted in the Helsinki peace agreement in

August 2005. Since then, Aceh has entered a period of relative peace.

The separatist movement in Aceh is not been the only post-Suharto conflict that Indonesia has

dealt with. Suppressed by the New Order’s military might, long-simmering tensions finally

11

broke out into overt conflict and destructive violence during the country’s transition to

democracy and decentralization. While conflicts occurred throughout the archipelago, some

regions were particularly affected by violence.

In the provinces of Maluku and North Maluku, tensions between religious and migrant groups

caused over 7,000 deaths between 1999 and 2002 (see Brown et al. (2005) for a conflict

overview). Populated by Muslims and Christians in approximately equal proportion, historical

inequalities stemmed from preferential treatment of Christians during Dutch colonialism. The

political, social, and economic dominance of Christians, however, began to erode under the

Islamization policies during the last decade of the New Order regime, when Muslims were

increasingly appointed to key positions in the civil service. Continued influx of mostly

Muslim migrants from Java and Sulawesi further challenged the fragile ethno-religious

balance, and resulting tensions over communal land and resources were aggravated by the

1997 economic crisis.

In January 1999, a fight between an Ambonese Christian bus driver and an immigrant Muslim

triggered a period of wide-spread inter-communal violence that quickly spread to other parts

of the province. The initial disinterest, or even active involvement, of security forces

contributed to an escalation of violence, and the conflict further intensified in mid-2000, when

several thousand fighters of the newly-founded Islamic militia organization Laksar Jihad

entered the region to support the Muslim cause. The central government’s build-up of security

forces along with Vice President Kalla’s efforts to mediate peace resulted in the Malino II

peace agreements in February 2002. While this peace is fragile and occasional violence

continues, conflict is considerably less intense (Varshney et al., 2009).

Comparable patterns of violence evolved in Sulawesi, located to the west of the Maluku

islands, and particularly in the province of Central Sulawesi. For similar reasons,4 tensions

between Christians and Muslims increased during the 1990s, gaving way to severe sectarian

violence in the post-Suharto period, with an estimated 1,000 killed and 100,000 displaced (see

e.g. Human Rights Watch (2002) for a conflict overview). With the absence of effective

security forces and functioning legal institutions, the conflict went through different phases

and intensified further with the arrival of Laskar Jihad fighters in mid-2001. The increased

4 Similarities in the underlying causes of the violence in the Moluccas and in Central Sulawesi include historical socio-economic imbalances, shifts in local power from Christians to Muslims as a result of Suharto’s ‘Islamic Turn’ in the 1990s, spontaneous and government-led in-migration of Muslims mainly from South Sulawesi and Java, as well economic uncertainty in the wake of the 1997 economic crisis.

12

presence of police and military personnel combined with the central government’s efforts to

mitigate the conflict finally led to the Malino peace declaration in December 2001. The

number of communal clashes has since declined substantially; incidents of violence, however,

still sporadically occur and peace remains vulnerable.

Conflicts reported outside these areas are spread across the archipelago, with more

pronounced levels of violence in the Nusa Tenggara islands, parts of Java and Central

Sumatra, as well as Papua. Underlying causes of these low intensity conflicts appear to be

manifold and related to local circumstances (Barron et al., 2009). Generally, a lack of

effective institutions and mechanisms of dispute resolution, local power struggles in the

process of decentralization, and conflicts related to land ownership and usage seem to be the

most prominent drivers of communal violence in Indonesia.

Growth Data

For the empirical analysis looking at the economic consequences of these conflicts, we draw

on two distinct data sources. Data on economic growth is provided by the Central Bureau of

Statistics, Badan Pusat Statistik (BPS). The Gross Regional Domestic Product (GRDP) data

are disaggregated at district5 and sectoral level. Table I in the Appendix provides an overview

of the different sectors, sub-sectors, and sub-categories. We distinguish between the primary,

secondary, and tertiary sector, as well as additionally disaggregating into the following nine

sub-sectors: agriculture, mining, manufacturing, energy, construction, commerce, transport

and communication, finance, and other services. For each category, we calculate the annual

rate of GDP growth based on the reported value added. In order to control for economic

spillover effects, we also calculate the average weighted growth rate in neighboring districts

in the current and preceding year.6 Apart from five missing observations,7 a balanced panel is

available.

It is important to note that the GDP figures are compiled separately by the BPS offices at

national, regional, and district levels, which is likely to result in varying qualities of the data

provided. Moreover, the BPS system does not guarantee that district and provincial figures

5 As of 2008, Indonesia is divided into 33 provinces, which are further divided into a total of 456 districts. 6 Districts are considered “neighboring” when sharing a common land border. 7 Control variables are not available from the district of Nias (North Sumatra). GDP growth rates can not be calculated for the newly founded districts of Kota Bengkulu (Bengkulu) and Kota Bau-Bau (South East Sulawesi) for 2002, as GDP data for 2001 is not available.

13

add up to provincial and national aggregates, respectively. Indeed, the sum of district GRDP

deviates, on average, by around three percent from the published provincial GRDP, with the

sum of district GRDP equaling 99 percent of the summed provincial GRDP figures. As these

deviations appear random across both regions affected and not affected by conflict, these

weaknesses in the data are unlikely to bias the analysis.

Still, in order to reduce noise in the data, we correct a few observations where the reported

production values deviate extremely from the overall trend. As these extreme growth rates are

distributed randomly across districts and sectors,8 correction of extreme sub-sectoral growth

rates should not involve systematic distortions of the data. We adjust “outliers” (growth rates

larger than 50 percent) in the following way: a) in cases where the current GDP value deviates

extremely from the general trend, the average annual growth rate from the previous two years

is used; b) sub-sectors with growth rates greater than 50 percent are excluded from the

calculation of sectoral (agriculture, industry, services) and overall growth rates. This

procedure results in the adjustment of 99 sub-sectoral growth rates in total, which corresponds

to less than one percent of the whole sample. Table 1a provides descriptive statistics for the

variables of economic growth. For robustness, we also run the analysis with the original data.

PODES Data

Control and conflict variables are derived from the Village Potential Statistics Survey

(Potensi Desa or PODES), a village-level census that is conducted three times per decade.

Also administered by BPS, PODES collects socio-economic information from all 69,000

Indonesian rural villages and urban neighborhoods. The survey is based on responses of the

village heads and includes a wide range of indicators, ranging from population characteristics

to infrastructure, economic activities, and social life.

We use three sequential rounds of PODES (2003, 2006, 2008), collected in August 2002, in

April 2005, and between April and May 2008, respectively, and aggregate the data at district

level, which is the administrative level for which economic growth data is available. A

challenge in constructing the database arises from Indonesia’s post-Suharto decentralization

legislation and the related formation of new districts. This process, known as pemekaran, led

to an increase in the number of districts, going from 376 in 2002, to 438 in 2005, and 465 in

8 For example, a district with very high reported growth in the transport sector shows moderate growth rates in all other sectors.

14

2008. We therefore realign the 2005 and 2008 data to match the 2002 district borders in order

to achieve a uniform dataset of 376 districts throughout.

While some of the standard controls in cross-country growth regressions are not available at

district level,9 the detailed village-level information provided by the PODES data allows us to

obtain a valid set of proxy variables. We use the share of villages with electricity and the

average educational attainment of village heads in a district to proxy for infrastructure

development and institutional quality, respectively. Further explanatory variables are the rate

of population growth, ethnic diversity, and natural disaster events. To control for the

devastating impact of the 2004 Indian Ocean tsunami that hit the northern parts of Sumatra,

we also include an indicator of tsunami-related physical destruction (UNORC, 2005). Basic

descriptive statistics for the control variables are presented in Table 1b.

In 2003 PODES, a section on conflict has been included for the first time. The village heads

report, inter alia, on incidences of conflict in the previous year, the number of conflict

fatalities, the amount of conflict-related material damage, and whether the conflict is resolved

or is still ongoing. Based on this information, a series of conflict variables at district level is

derived. In order to obtain comparable indicators of conflict intensity, fatalities and material

damage are set in proportion to the district’s total population and total GDP, respectively.

Conflict spillover variables are computed by the ratio of the total number of fatalities (total

amount of material damage) in all neighboring districts to the total population (total GDP) of

all neighboring districts. Unlike binary indicators used in related cross-country analyses, this

approach allows for the consideration of the economic impact of conflict intensity in

neighboring regions.

Furthermore, we aim to address potentially varying growth effects in different phases of

violent conflict. However, as the PODES data covers conflict in the respective previous year

and is conducted three times per decade, the information on the occurrence of violent conflict

is not continuous. Although this prevents us from understanding the whole course of the

conflict, village heads do indicate whether conflict is resolved or still ongoing. This

information allows us to distinguish between active conflicts and observations from an early

post-conflict situation, and we split the indicators of relative conflict intensity accordingly.

9 For instance, no measure of human capital is included, as data on literacy or school enrolment rates are not available for 2008. Given the relatively short time period covered by the analysis, we generally do not expect these indicators to change substantially. In fact, the variance decomposition of available literacy data from 2000 to 2006 reveals low within district variations when compared to variation across districts.

15

Table 1c provides descriptive statistics for all conflict variables that we use in the analysis.

One concern we face is the issue of data reliability. Comparably low levels of conflict

reported by PODES raise concerns of misreporting by local authorities. When conflict is

perceived as the failure of local leaders and when respondents have doubts about the purpose

of the survey, violence might not be accurately reported. Comparisons with other quantitative

surveys and qualitative fieldwork indeed suggest that conflict is significantly underreported

by PODES.10 However, these comparisons also show such underreporting to be of similar

magnitude across different regions and hence systemic, i.e. not biased by local characteristics.

PODES data is therefore assumed to provide accurate information on relative levels of

conflict both across districts and over time. Keeping in mind that the potential underreporting

would result in an attenuation bias, we expect rather conservative estimates of the economic

impact of violent conflict.

ECONOMETRIC APPROACH

Building on standard growth regressions, the starting point of our empirical analysis is the

following equation:

git = Citα + Xitβ + εit , (1)

where git is the total or sectoral growth rate of real GDP in district i and year t, Cit are

measures of conflict, Xit is a matrix of other covariates that determine economic growth, and

εit is the unobservable error term. The coefficient of interest is α, which captures the effects of

conflict on GDP growth. Previous macro-level studies on the economic impacts of violent

conflict have in large part relied on pooled OLS estimation (Bozzoli et al., 2008). In this

setting, however, the underlying assumption of independent and identically distributed errors

ε is likely to be violated.

When factors that affect a society’s economic outcomes and its vulnerability to violent

conflict at the same time are omitted, OLS estimates are inconsistent. Cultural attitudes, for

instance, are hard to measure, but potentially drive both growth and conflicts. Given the panel

structure of the data, one way to deal with such unobserved heterogeneity is the introduction

of district fixed effects (FE) to control for those district characteristics that did not change

10 Barron and Sharpe (2008) monitor conflicts reported in local newspapers in the provinces of East Java and Nusa Tenggara Timur (NTT) during the period 2001-2003; Barron et al. (2004) conduct village-level case studies in the same provinces for comparison with 2003 PODES.

16

over the 2002-2008 period. This captures the underlying cultural values, as well as other time-

invariant or long term, slow changing, growth determinants, such as initial wealth or

geographic and climatic conditions.

The inclusion of time dummies further accounts for overall macroeconomic trends that affect

all districts alike. Therefore, only additional explanatory variables that vary both with district

and over time need to be included for the analysis of short term and within-district growth

dynamics. The variance decomposition for both the growth and conflict variables reveals

larger variations over time (within variation) than across districts (between variation), which

confirms the accuracy and necessity of the FE model, in particular, for our focus on the role of

violent conflicts.

As every empirical investigation of the conflict-growth nexus, we face concerns of reverse

causality. That is, (low) economic growth is not only a result, but also a potential cause of

violent conflict, leading to inconsistent parameter estimates. Ideally, we would like to

instrument for the endogenous variable(s) to disentangle simultaneous causation. Finding

suitable “instruments”, however, which predict the incidence or even intensity of violent

conflict and are uncorrelated with the error term, is extremely difficult. Instruments for

conflict proposed by the literature include agricultural growth, urban population, and tropical

location (Kang and Meernik, 2005). Further, the size of the population and the geophysical

environment may considered as potential IV candidates.11

All of these instruments seem neither valid nor relevant. While the required exogeneity with

respect to economic growth is disputable, its explanatory power in predicting violence appears

even more problematic. The outbreak and the course of violent conflict is, in general, highly

complex, and attempting to capture these processes with only one or a few proxy variables

inevitably runs the risk of oversimplification and misframing.12 As convincing instruments are

hardly imaginable, we forego the IV approach, and investigate potential reverse causality

from growth to conflict for the Indonesian context in more detail.

11 Collier and Hoeffler (2004) find population size to be an important determinant of war, with more populous countries being more prone to conflict. Fearon and Laitin (2003) stress that mountainous terrain (similar: forest coverage) can facilitate civil wars, as rough terrain provides sanctuary to rebels. 12 Potential reverse effects from growth to conflict are likewise contested. In an innovative approach, Miguel et al. (2004) use abrupt declines in rainfall in Sub-Saharan Africa, where economies primarily rely on rain-fed agriculture, to instrument for economic growth. While they find a negative impact of economic growth on conflict incidence, this result, however, seems not to hold when a more appropriate sub-sample of states with conflict on their own territory is used – as opposed to the inclusion of states that send troops to conflicts in other states (Jensen and Gleditsch, 2009).

17

The underlying reasons for the conflicts in Indonesia at the turn of the millennium are

multifarious and even vary across the affected regions. While the conflict in Aceh was driven

by the long quest for (more) independence, the widespread communal violence rose from

latent tensions related to horizontal inequalities and large-scale migration, the lack of effective

security and justice institutions, and local-level struggles for economic power in the process

of decentralization. Economic uncertainties in the wake of the financial crisis certainly

contributed to the 1999 outbreak of violence, and disappointing economic developments

thereafter are likely to have affected people’s decision to engage in conflict.

However, focusing on the 2002 PODES wave, GDP growth seems not to be among the most

obvious drivers of reported violence: first, the conflicts reported by the village heads represent

the September 2001 to August 2002 period. The analysis therefore captures the impact of

conflict on slightly lagged growth. This is especially true for Central Sulawesi and the

Maluku islands, where increased levels of conflict in 2001 were followed by the Malino peace

agreements signed in December 2001 and February 2002, respectively.

Second, current economic growth is unlikely to have significantly influenced the intensity of

violence in the case of these mostly ongoing conflicts. Rather, high levels of violence in Aceh

between 1999 and 2002 were related to a brutalization of the conflict aimed at influencing the

outcomes of the negotiations on autonomy. Intensified clashes between the Indonesian

military and the GAM were increasingly accompanied by atrocities against civilians.

Similarly, provocations by the military and the massive arrival of non-local Laskar Jihad

fighters fueled the intensification of the conflicts in the Maluku islands and Central Sulawesi.

Hence, conflict intensity and frequency were mainly driven by outside interventions, rather

than by short-term evolutions of the local economy. While being aware of potential

distortions due to reverse causality and not claiming conflict to be exogenous to growth, we

believe that causation in this setting strongly runs from conflict to growth, rather than the

opposite way. The estimates from the fixed effects regressions are therefore assumed to

reflect the growth impact of violent conflict in a fairly accurate way.

EMPIRICAL RESULTS

Descriptive Statistics

Figure 1 displays the evolution of sectoral and overall GDP growth rates in Indonesia for the

2001-2008 period. Overall, weighted annual growth appears relatively stable at around five

18

percent and increases slightly over time. The service sector, which accounts for 45 percent of

total GDP in 2008 (see Table I in the Appendix for the sectoral shares of total GDP in 2000

and 2008), is the main driver of the Indonesian economy with average annual growth rates

above six percent. Above-average growth rates are also found in the industrial sector, with

manufacturing being the most important sub-sector and accounting for one quarter of total

GDP. Relatively low growth is observed for the primary sector. It still accounts for 26 percent

of total rural production in 2008, with its share only decreasing slowly since 2000.

Table 2 summarizes the occurrence of conflict over time, as reported by village heads. We

observe a substantial decrease in both the spread and the intensity of violence in 2005 and

2008, in comparison with 2002.13 While 7.2 percent of all villages and neighborhoods report

conflict in 2002, this figure is halved in 2005 and 2008. Even more striking is the decline in

conflict-related human losses and physical destruction over time: fatalities and material

damage in 2005 and 2008 each amount to only approximately one tenth of the 2002 figures,

which documents Indonesia’s move toward peace in recent years.

With most of the violence in 2002, we take a closer look at the spatial distribution of violence

in this year. Figure 2a maps conflict in terms of fatalities per capita, Figure 2b shows conflict

intensities in terms of material damage relative to district GDP. Conflict-related fatalities are

relatively concentrated and particularly prevalent in the entire province of Aceh, as well as on

the islands of Maluku, North Maluku, and Sulawesi. High levels of physical destruction are,

in addition, observed in the province of Jambi and on the islands of Nusa Tenggara, while

almost no material damage is reported from the western coast of Aceh.

Table 3 lists the districts in the sample which are most affected by violence, with observations

almost exclusively from 2002 and from rural areas. In fact, only ten percent of overall

fatalities and seven percent of overall material damage are reported from urban districts,

which make up for around 20 percent of the total population in 2002. The figures also confirm

the differences in the distribution of human losses and destroyed physical capital: a

substantial share of those districts with high numbers of fatalities are comparatively little

affected in terms of physical destruction, and vice versa (correlation between reported

material destruction and fatalities: 0.66). We use this variation to analyze potentially varied

economic impacts of different forms of violence.

13 Village heads were asked for incidents of conflict in the previous year, which hence relates to conflicts in the periods September 2001-August 2002, June 2004-May 2005, and May/June 2007-April/May 2008. For simplicity, we refer to conflicts in 2002, 2005, and 2008, respectively.

19

In 2002, the main conflicts in Aceh, Maluku, and Sulawesi account for 89 percent of total

fatalities (per capita) and 67 percent of total material damage (relative to the district’s GDP)

reported by village heads. Outside these regions, mostly low intensity forms of violence are

reported from various parts of the country. However, it has to be noted that the majority of

districts were hardly affected by any conflict, even in 2002, when 81 percent of the districts

reported less than five conflict-related fatalities.

Regression Results

Table 4 presents the FE regression results for the main sectors. The dependent variable is the

annual growth rate of district GDP. We run separate regressions for agriculture, industry,

service, and total GDP growth. Some of the control variables are not significant. Unlike

previous cross-country studies, we fail to find significant impacts of population growth, ethnic

diversity, or infrastructure development (here: availability of electricity) on economic growth.

A short survey period, the inclusion of district fixed effects, and the relative invariability of

the variables in the short term are most likely explanations. A significantly positive impact on

industry and overall growth is found for average educational attainment of village heads,

which is supposed to measure improvements in the quality of institutions.

The two indicators of natural disasters yield mixed results. While the incidence of any kind of

disaster (earthquake, landslide, and flood) in the years prior to the survey has no significant

effect, 14 the impact of the 2004 tsunami on subsequent economic growth in the affected

regions is substantially negative, and particularly so for agriculture. No spillover effects from

current and recent economic growth in a district’s vicinity are found, and time dummies

confirm overall greater economic growth in 2005 and 2008, in comparison to 2002.

We include three indicators of violent conflict: the amount of material damage relative to total

district GDP, the number of fatalities relative to district population size, and an indicator of

conflict intensity in neighboring districts.15 Results indicate a substantial and significantly

negative impact of conflict-related physical destruction on industry, service, and total GDP

growth. These estimates translate into an approximate 11 percentage points decrease in

industrial growth, and a 5 percent age points decrease in service sector growth for the two

14 The lack of significance is likely to be caused by the long, three-year period under consideration, the relatively frequent occurrence of floods in Indonesia, and missing information on the scale of destruction. 15 Here, the conflict spillover variable is based on relative material damage in neighboring districts. An indicator based on the number of fatalities is used in later regressions.

20

most affected districts of Aceh Timur and Poso (Central Sulawesi). A district affected by the

average 2002 conflict intensity would, accordingly, expect a decrease in growth by 0.2

(industry) and 0.1 (services) percent age points, respectively.

The conflict coefficients for agriculture are small in magnitude and insignificant, which seems

to confirm the primary sector’s relative immunity to violent conflict. So far, we do not find

any spillover effects from conflicts in neighboring districts on domestic growth. Likewise, the

coefficients for conflict fatalities are not significant, with the marginal exception of the

service sector. Keeping potential multicollinearity among the conflict variables in mind,16 this

points to a more detrimental economic impact of physical destruction, compared to human

losses. Before having a closer look into these varied effects for different forms of violence, we

turn to the sub-sector regression results to investigate the particular economic effects of

violent conflict in more detail.

Table 5 presents similar growth regressions for eight sub-categories of the industry and

service sectors. The agricultural sector is omitted here, as no further disaggregation is

available. While the same set of control variables as before are used, we focus on the growth

effects of conflict-related material damage and drop the simultaneous inclusion of the

insignificant fatality indicator. Some of the previously insignificant control variables now

become significant for specific branches, such as infrastructural development for finance,

transport, and communication growth. Natural disasters have particular impact on the

industrial sector, while services seem less severely affected. Neither economic growth nor

conflicts in neighboring districts significantly influence domestic outcomes.

The conflict coefficients confirm our expectations and existing empirical evidence from

previous sector-specific analyses. We find particularly strong negative conflict effects for

financial services, transport/communication branches, and manufacturing industries.

Substantial and significantly negative coefficients are also reported for construction industries

and the commerce sector, which includes retail, hotels, and restaurants. The mining and

energy sectors, i.e. electricity, gas, and water supply, appear unaffected. The only sub-

category of the tertiary sector with an insignificant conflict coefficient is the area of

‘services’. As government expenditures for administration and defense constitute the bulk of

this category, maintained or increased security spending might be a potential driver behind

this result.

16 The Variance Inflation Factor (VIF), a simple measure to identify multicollinearity, is calculated and the results (VIF<2 for all explanatory variables) mitigate concerns of multicollinearity.

21

In the next step, we exploit some of the growth dynamics of violent conflict. Ceteris paribus,

we split the indicator of relative physical destruction into ongoing and recently resolved

conflicts. Table 6 presents the results from separate regressions on total, sectoral, and sub-

sectoral GDP growth. As the general regression setup has not changed, the estimates for the

control variables resemble those obtained in previous regressions and are therefore not

discussed further here.

Focusing on the estimated growth effects of ongoing and recent conflicts, we find clear-cut

differences. For most sectors, results point to a considerably stronger decline in growth rates

during ongoing conflicts than indicated by the singular indicator of conflict previously used.

In particular, the construction and transport/communication sectors seem to suffer most

during episodes of violence. Collapsing demand for buildings and infrastructure combined

with an overall decrease in business transactions potentially deepens the crisis in these sectors

during times of conflict. A relatively strong negative growth impact of conflict is also found

for manufacturing and commerce industries, which is in line with both Collier’s (1999)

concept of war vulnerability and existing evidence on the effects of violence on tourism and

hospitality industries.

Turning to the growth impact for resolved conflicts, the estimates are by far less substantial

and significant. The only exception is the services sub-category: including, inter alia,

government defense expenditures, this is possibly related to a post-conflict cutback in public

spending on security. Marginally negative impacts of resolved conflicts are still found for the

financial sector and manufacturing, which are assumed to be particularly dependent on a

stable market environment. By contrast, the even insignificantly positive ‘post-war’

coefficients for construction and transport/communication industries seem related to public

and private investment in reconstruction activities.

Robustness Tests

A series of alternative sub-samples and conflict coefficients are employed to test the

robustness of our findings. As the regression set-up and the control variables remain

unchanged, we only present the estimates for the conflict coefficients in what follows.17

First, the sample is restricted to rural areas, where, by far, most of the violence took place.

17 The complete set of regression results is available from the authors upon request.

22

The results, presented in Table 7a, are close to those for the whole sample: growth rates in the

fields of transport/communication, construction, finance, and manufacturing are substantially

affected during ongoing conflict. We find a marginally negative impact on commerce growth,

while the positive post-conflict trend for the construction sector is strengthened. In addition,

the results indicate negative conflict spillover effects on finance and transport/communication.

While the transport sector relies, almost by definition, on cross-district activities, these results

also suggest that trust and transactions in domestic financial markets are affected by

neighboring conflicts.

Second, we exclude observations from the province of Aceh, to see to what extent our

findings are driven by the separatist conflict in this region (Table 7b). Once Aceh is excluded,

the negative effects of conflict on the tertiary sector turns become less significant and are

partly absorbed by the conflict spillover variables. The magnitude of the conflict coefficients,

however, remains largely unchanged. Interestingly enough, the immediate recovery of the

construction sector after the end of conflict is strongly supported here. This effect becomes

even clearer when we instead exclude the islands of Maluku and Sulawesi and the associated

ethno-communal conflicts (Table 7c). In line with the Collier’s (2009) characterization of the

recovery process, we find the immediate post-conflict recovery efforts to center on the

reconstruction of critical infrastructure. This trend is also supported by the substantially

positive post-conflict growth estimate for the energy sector.

Third, instead of conflict-related material damage, we use alternative indicators of conflict.

Table 7d presents the estimates for the indicators of conflict fatalities. Interestingly enough,

the distinction between ongoing and resolved conflicts reveal substantial and significantly

negative effects of human losses for ongoing conflicts. Consistent with previous findings for

conflict-related material damage, finance, transport/communication, construction, and

manufacturing are the most negatively affected. The results for resolved conflicts further

confirm the impression of distinctly different economic impacts at different stages of violent

conflict.

A second alternative indicator of conflict is the proportion of villages in a district that report

incidences of conflict. Consistently, the coefficients are highly insignificant (Table 7e). As

this indicator does not take into account the severity of violence, the mere existence of

conflict seems to have no a priori impact on the economy. As claimed by related cross-

country studies, the intensity of conflict decisively determines the impact on the economy.

This is also confirmed when using absolute numbers of material damage and fatalities instead

23

of relative indicators of physical destruction and human losses (Tables 7f and 7g). The results

are largely in line with our previous findings and underscore the particular conflict

vulnerability of the service sector.

Finally, we repeat the main regressions with the original GDP data, without correcting for

outliers in sub-sectoral growth rates. Results for the whole sample and the sub-sample of rural

areas are presented in Table II.a and II.b of the Appendix, and confirm our main findings.

Again, we repeat the analysis with the indicator of conflict fatalities (Table II.c in the

Appendix) and obtain similar results.

CONCLUSION

The overall negative effect of violent conflict on economic outcomes is confirmed by a

growing amount of literature. However, relatively little is known on sector-specific impacts of

violence, as disaggregated GDP data is scarcely available from conflict-ridden countries. In

this paper, we build on detailed, district-level, GDP data from Indonesia and on nationwide

data on the occurrence of conflicts, which allows us to provide quantitative evidence on the

vulnerability to conflict across different economic activities.

We find significantly different effects of violent conflict across sectors. As proposed by

Collier (1999), industries dependent on either capital or transaction are most vulnerable to

conflict. In particular, this applies to manufacturing, finance, transport and communication, as

well as the retail and hospitality industries. The construction sector suffers substantially

during active conflicts, while a slightly positive trend is found in the early post-conflict

period. Other sectors’ recovery appears less rapid; both manufacturing industries and the

finance sector seem especially reliant on a lasting peace.

As we do not control for saving rates, reduced investment is a potential driver of the economic

downturn in times of conflict. Overall, the analysis points to only temporary, short term

impact of violent conflict on economic growth. A limitation of this study is the intermittent

character of the conflict data. Observations on conflict before and after the respective survey

years are not available, which prevents us from conducting a more detailed analysis of the

growth dynamics during conflict and in post-conflict times. Nevertheless, the findings point to

substantial differences in the sector-specific impacts of violent conflict, with potentially

important implications for the reconstruction process and future development.

The state’s and other actors’ efforts to foster economic recovery in the post-conflict period

24

should be selective in the sense that they particularly target the most affected branches. The

rebuilding of infrastructure, such as transportation and communication networks, seems

crucial especially for manufacturing industries and logistics services. Further, restoring peace

and security probably contributes most to the recovery of the financial and tourism sectors.

Given the scarcity of public funds in the aftermath of conflict, consideration of the local

economy’s structure and the varying needs across branches should guide and contribute to an

efficient use of resources.

25

REFERENCES

Abadie, A., Gardeazabal, J., 2003. The Economic Costs of Conflict: A Case Study of the Basque Country. American Economic Review 93 (1), 113-32.

Acemoglu, D., Johnson, S., Robinson, J., 2004. Institutions as the Fundamental Cause of Long-Run Growth, in Aghion, P., Durlauf, S. (Eds.), Handbook of Economic Growth, Amsterdam.

Acemoglu, D., Johnson, S., Robinson, J., 2001. Colonial Origins of Comparative Development: An Empirical Investigation. American Economic Review 91, 1369-401.

Alesina, A., Baqir, R., Easterly, W., 1999. Public Goods and Ethnic Divisions. Quarterly Journal of Economics 114, 1243-84.

Annan, J., Blattman, C., Horton, R., 2006. The State of Youth and Youth Protection in Northern Uganda: Findings from the Survey of War Affected Youth. UNICEF.

Arunatilake, N., Jayasuriya, S., Kelegama, S., 2001. The Economic Cost of the War in Sri Lanka. World Development 29 (9), 1483-1500.

Balaguer, J., Cantavella-Jordá, M., 2002. Tourism as a Long-run Economic Growth Factor: The Spanish Case. Applied Economics 34 (7), 877-84.

Barro, R.J., McCleary, R.M., 2003. Religion and Economic Growth. American Sociological Review 68 (5), 760-81.

Barron, P., Kaiser, K., Pradhan, M., 2009. Understanding Variations in Local Conflict: Evidence and Implications from Indonesia. World Development 37 (3), 698-713.

Barron, P., Kaiser K., Pradhan, M., 2004. Local Conflict in Indonesia: Measuring Incidence and Identifying Patterns. World Bank Policy Research Working Paper 3384.

Barron, P., Sharpe, J., 2008. Local Conflict in Post-Suharto Indonesia: Understanding Variations in Violence Levels and Forms through Local Newspapers. Journal of East Asian Studies 8, 395-424.

Blattman, C., Miguel, E., 2010. Civil War. Journal of Economic Literature 48 (1), 3-57.

Blomberg, S.B., Hess, G.D., 2006. How Much Does Violence Tax Trade?. The Review of Economics and Statistics 88 (4), 599-612.

Bodea, C., Elbadawi, I.A., 2008. Political Violence and Economic Growth. World Bank Policy Research Working Paper 4692.

Bos, J.W.B., Economidou, C., Koetter, M., Kolari, J.W., 2010. Do All Countries Grow Alike?. Journal of Development Economics 91, 113-27.

Bosma, N., Stam E., Schutjens, V., 2010. Creative Destruction and Regional Productivity Growth: Evidence from the Dutch Manufacturing and Services Industries. Small Business Economics, forthcoming.

Bosworth, B.P., Collins, S.M., 2003. The Empirics of Growth: An Update. Brookings Papers on Economic Activity 2, 113-206.

Bozzoli, C., Brück, T., Drautzburg, T., Sottsas, S., 2008. Economic Costs of Mass Violent Conflict. Politikberatung kompakt 42, DIW Berlin.

Brown, G., Wilson, C., Hadi, S., 2005. Overcoming Violent Conflict: Peace and Development Analysis in Maluku and North Maluku. CPRU-UNDP, LIPI and BAPPENAS.

26

Brück, T., 1997. Macroeconomic Effects of the War in Mozambique. Queen Elizabeth House Working Paper Series 11, University of Oxford International Development Centre.

Cerra, V., Saxena, S.C., 2008. Growth Dynamics: The Myth of Economic Recovery. American Economic Review 98 (1), 439-57.

Collier, P., 1999. On the Economic Consequences of Civil War. Oxford Economic Papers 51, 168-83.

Collier, P., 2009. Post-conflict Recovery: How Should Strategies Be Distinctive?. Journal of African Economies 18, AERC Supplement 1, 99-131.

Collier, P., Hoeffler, A., 2004. Greed and Grievance in CivilWar. Oxford Economic Papers 56 (4), 563-95.

Deininger, K., 2003. Causes and Consequences of Civil Strife: Micro-Level Evidence from Uganda. Oxford Economic Papers 55, 579-606.

De Groot, O., 2010. The Spill-Over Effects of Conflict on Economic Growth in Neighboring Countries in Africa. forthcoming in Defence and Peace Economics.

Depetris Chauvin, N.M., Rohner, D., 2009. The Effects of Conflict on the Structure of the Economy. Dubai School of Government Working Paper 09-10.

Durlauf, S.N., Johnson, P., Temple, J., 2005. Growth Econometrics, in P. Aghion and S.N. Durlauf (Eds.), Handbook of Economic Growth, Amsterdam.

Easterly, W., Levine, R., 2001. It’s Not Factor Accumulation: Stylized Facts and Growth Models. World Bank Economic Review 15 (2), 177-219.

Esfahani, H.S., Ramírez, M.T., 2003. Institutions, Infrastructure, and Economic Growth. Journal of Development Economics 70, 443-77.

FAO, 1997. FAO/WFP Crop and Food Supply Assessment Mission to Rwanda 1 July 1997. Special Reports and Alerts (GIEWS)’.

Fearon, J.D., Laitin, D.D., 2003. Ethnicity, Insurgency, and Civil War. American Political Science Review 97, 75-90.

Fielding, D., 2000. Counting the Cost of the Intifada: Consumption, Saving and Political Instability in Israel. Discussion Paper in Economics 00/8, University of Leicester.

Fielding, D., 2004. How Does Violent Conflict Affect Investment Location Decisions? Evidence from Israel during the Intifada. Journal of Peace Research 41 (4), 465-84.

Frankel, J., Romer, D., 1999. Does Trade Cause Growth?. American Economic Review 89 (3), 379-99.

Garcia, J.G., Soelistianingsih, L., 1998. Why Do Differences in Provincial Incomes Persist in Indonesia?. Bulletin of Indonesian Economic Studies 34 (1), 95-120.

Ghobarah, H.A., Huth, P., Russett, B., 2003. Civil Wars Kill and Maim People - Long After the Shooting Stops. American Political Science Review 97 (2), 189-202.

Glick, R., Taylor, A.M., 2005. Collateral Damage: Trade Disruption and the Economic Impact of War. CEPR Discussion Paper 5209.

Gupta, S., Clements, B., Bhattacharya, R., Chakravarti, S., 2004. Fiscal Consequences of Armed Conflict and Terrorism in Low- and Middle-Income Countries. European Journal of Political Economy 20, 403-21.

27

Human Rights Watch, 2002. Breakdown: Four Years of Communal Violence in Central Sulawesi. C1409, New York.

Imai, K., Weinstein, J., 2000. Measuring the Economic Impact of Civil War. CID Working Paper 51, Harvard University.

Jensen, P.S., Gleditsch, K.S., 2009. Rain, Growth, and Civil War: The Importance of Location. Defense and Peace Economics 20 (5), 359-72.

Kang, S., Meernik, J., 2005. Civil War Destruction and the Prospects for Economic Growth. Journal of Politics 67 (1), 88-109.

Knight, M., Loayza, N., Villanueva, D., 1996. The Peace Dividend: Military Spending Cuts and Economic Growth. World Bank Policy Research Working Paper 1577.

Koubi, V., 2005. War and Economic Performance. Journal of Peace Research 42 (1), 67-82.

Lai, B., Thyne, C., 2007. The Effects of Civil War on Education 1980-97. Journal of Peace Research 44 (3), 277-92.

Lopez, H., Wodon, Q., 2005. The Economic Impact of Armed Conflict in Rwanda. Journal of African Economies 14 (4), 586-602.

Masters, W., McMillan, M., 2001. Climate and Scale in Economic Growth. Journal of Economic Growth 6 (3), 167-86.

Mawardi, S., 2006. Community Synthesis Report: Kampung Pisang, Ternate, North Maluku, Indonesia. The SMERU Research Institute, Jakarta.

McCulloch, N., Sjahrir, B.S., 2008. Endowments, Location or Luck? Evaluating the Determinants of Sub-National Growth in Decentralized Indonesia. World Bank Policy Research Working Paper 4769.

Miguel, E., Satyantah, S., Sergenti, E., 2004. Economic Shocks and Civil Conflict: An Instrumental Variables Approach. Journal of Political Economy 112 (4), 725-753.

Murdoch, J.C., Sandler, T., 2002. Economic Growth, Civil Wars, and Spatial Spillovers. Journal of Conflict Resolution 46 (1), 91-110.

Neumayer, E., 2004. The Impact of Political Violence on Tourism: Dynamic Cross-National Estimation. Journal of Conflict Resolution 48 (2), 259-81.

Organski, A.F.K., Kugler, J., 1977. The Costs of Major Wars: The Phoenix Factor. American Political Science Review 71 (4), 1347-66.

Psacharapoulos, G., Patrinos, H.A., 2004. Returns to Investments in Education: A Further Update. Education Economics 12 (2), 111-34.

Rao, B.B., Corray, A., 2009. How Useful is Growth Literature for Policies in the Developing Countries?. MPRA Paper 14573.

Reid, A., (Ed.), 2006. Verandah of Violence: The Background to the Aceh Problem. Singapore University Press.

Ross, M.L., 2004. What Do We Know About Natural Resources and Civil War?. Journal of Peace Research 41 (3), 337-56.

Sapio, S., Thoma, G.L.D., 2006. The Growth of Industrial Sectors: Theoretical Insights and Empirical Evidence from U.S. Manufacturing. LEM Working Paper 2006/09.

Solow, R.M., 1956. A Contribution to the Theory of Economic Growth. Quarterly Journal of Economics 70, 65-94.

28

29

Svensson, J., 1998. Investment, Property Rights and Political Stability: Theory and Evidence. European Economic Review 42, 1317-41.

Verpoorten, M., 2009. Household Coping in War- and Peacetime: Cattle Sales in Rwanda, 1991–2001. Journal of Development Economics 88 (1), 67-86.

Tadjoeddin, M.Z., Murshed, M.M., 2007. Socio-Economic Determinants of Everyday Violence in Indonesia: An Empirical Investigation of Javenese Districts, 1994-2003. Journal of Peace Research 44 (6), 689-709.

Teodosijević, S.B., 2003. Armed Conflicts and Food Security. ESA Working Paper 03-11.

Office of the UN Recovery Coordinator for Aceh and Nias (UNORC), 2005. Tsunami Recovery Status Report. Banda Aceh.

Varshney, A., Barron, P., Jaffrey, S., Palmer, B., Rasyid, I., 2009. Conflict After the Big Wave: Early Evidence from Maluku and North Maluku. Presentation at the World Bank, Jakarta, 12th August 2009.

Table 1: Summary Statistics

Variable n Mean Std. Dev. Min Max

District GDP growth: Agriculture 1120 3.45 4.55 -40.49 37.33

District GDP growth: Mining 1036 4.76 8.10 -48.49 48.32

District GDP growth: Manufacturing 1118 4.42 5.09 -48.62 48.36

District GDP growth: Energy 1115 6.70 7.32 -48.48 47.70

District GDP growth: Construction 1115 6.69 6.55 -40.68 47.95

District GDP growth: Industry 1123 4.78 5.21 -37.51 42.39

District GDP growth: Commerce 1121 6.04 4.57 -43.13 43.54

District GDP growth: Transport and Communication 1117 6.85 5.10 -27.69 46.59

District GDP growth: Finance 1104 6.86 6.80 -40.48 46.72

District GDP growth: Services (sub-category) 1110 5.66 4.85 -49.33 45.22

District GDP growth: Services 1123 6.06 3.69 -41.56 34.81

District GDP growth: Total 1123 4.75 3.45 -33.57 35.91

Spillovers: weighted* average agricultural growth 1123 3.65 2.20 -14.95 11.13

Spillovers: weighted average industrial growth 1123 4.13 4.30 -31.47 31.57

Spillovers: weighted average service sector growth 1123 5.84 2.68 -21.43 17.95

a. E

cono

mic

Gro

wth

Spillovers: weighted average total GDP growth 1123 4.23 2.74 -26.23 21.97

Share of villages where electricity is available 1123 94.78 13.07 0.74 100

Average level of education of village heads (1-6) 1123 4.30 0.71 1.45 6

Population growth rate 1123 2.00 2.32 -4.83 14.77

Share of villages with more than one ethnicity 1123 72.74 23.39 2.42 100

Natural disaster in the last three years 1123 0.376 0.238 0 1

b. C

ontr

ols

Tsunami: number of new houses required (per 1,000 Inh.) 1123 0.046 0.614 0 13.66

Share of villages affected by conflict 1123 0.044 0.062 0 0.72

Share of villages affected by violent conflict** 1123 0.029 0.040 0 0.38

Fatalities per 1,000 inhabitants 1123 0.015 0.138 0 2.86

Fatalities per 1,000 inhabitants (resolved conflicts) 1123 0.008 0.080 0 1.92

Fatalities per 1,000 inhabitants (ongoing conflicts) 1123 0.007 0.095 0 2.78

Material damage as .01 %-share of GDP 1123 0.050 0.342 0 6.98

Material damage as .01 %-share of GDP (resolved conflicts) 1123 0.031 0.201 0 5.19

Material damage as .01 %-share of GDP (ongoing conflicts) 1123 0.019 0.225 0 6.64

Spillovers: fatalities per 1,000 inhabitants 1123 0.017 0.109 0 1.97

c. C

onfl

ict

Spillovers: material damage as .01 %-share of GDP 1123 0.041 0.224 0 4.93 * The average neighboring growth rate ng for sector s in district i is equal to the sum of the growth rates gn,s in the neighboring districts

n=1,…,mi, weighted by the neighbor’s economic size GDPn,s relative to the economic size of all the mi surrounding districts:

,, ,

1 ,1

( )i

i

mn s

i s n s mn n sn

GDPng g

GDP

** Violent conflict is defined as conflict that involves either human or physical losses.

30

31

Table 2: Violent Conflict – Descriptive Statistics

CONFLICT INDICES 2002 2005 2008

Total

Share of villages affected by conflict (%) 7.2 2.7 3.2

Share of villages affected by violent conflict (%) 4.4 2.1 2.3

Number of fatalities 4,858 276 335

Material damage (million of Rupiah) 740,560 97,742 8,476

District Means

Number of fatalities 13.0 0.7 0.9

Material damage (million of Rupiah) 2,054 260 23

Fatalities per 1,000 inhabitants 0.042 0.002 0.002

Material damage as .01 %-share of GDP 0.132 0.016 0.002

Table 3: Districts most affected by violence

(A) Province District Fatalities

per Capita (B) Province District

Mat. Damage (Share GDP)

(1) NAD Aceh Timur 2.86 (1) NAD Aceh Timur 6.98

(8) NAD Aceh Tengah 2.08 (4) Sulawesi Tengah Poso 6.65

(22) Maluku Utara Halmahera Barat 1.97 (16) Maluku Pulau Buru 3.99

(2) Sulawesi Tengah Poso 1.39 (11) NAD Kota Langsa 1.51

(91) Maluku Kota Ambon 1.24 (244) Jambi Kerinci 1.48

(71) NAD Aceh Selatan 0.77 (47) Sulawesi Selatan Mamuju 1.25

(15) NAD Aceh Barat Daya 0.63 (220) Jambi Merangin 1.24

(249) NAD Aceh Besar 0.34 (2) NAD Aceh Tengah 1.21

(12) NTT Rote Ndao 0.30 (29) NTB Dompu 1.13

(95) NAD Aceh Utara 0.26 (44) Maluku Maluku Tenggara 1.07

(4) NAD Kota Langsa 0.25 (411) Sulawesi Selatan Kota Palopo 1.07

(346) NAD Nagan Raya 0.22 (9) NTT Rote Ndao 1.02

(346) NAD Aceh Barat 0.20 (411) Jambi Bungo 1.01

(47) Kalimantan Tengah Kotawaringin Barat 0.18 (21) NAD Pidie 1.01

(346) Papua Puncak Jaya 0.17 (7) NAD Aceh Barat Daya 1.00

(3) Maluku Pulau Buru 0.17 (98) Maluku Maluku Tengah 0.96

(32) Maluku Maluku Tengah 0.16 (23) Sulawesi Selatan Luwu 0.88

(234) NAD Kota Lhokseumawe 0.15 (64) NAD Aceh Tamiang 0.84

(80) NAD Bireuen 0.13 (343) NTB Lombok Tengah 0.81

(103) NAD Aceh Singkil 0.12 (47) Sulawesi Tenggara Muna 0.75

Fatalities: conflict-related fatalities per 1,000 inhabitants. Material damage as .01%-Share of Total District GDP.

All observations from 2002 (observations from 2005 underlined). Urban districts in italic.

In parentheses: ‘rank’ of the district in terms of material damage (A) and fatalities (B), respectively.

Figure 1: Growth Trends 2001 - 2008

Growth (Total)

0%

1%

2%

3%

4%

5%

6%

7%

8%

9%

2001 2002 2003 2004 2005 2006 2007 2008

Agriculture Total Industry Services

Figure 2: Distribution of Violence in 2002 [intended for color reproduction on the web and in print]

a) Fatalities per Capita

b) Material Damage (as Share of Total District GDP)

32

33

Table 4: Base FE Regression Results

(1) (2) (3) (4) Dependent Variable: GDP Growth Agricult. Industry Services Total

0.06 0.01 -0.05 -0.08 Population Growth Rate

(0.08) (0.09) (0.06) (0.07)

-0.05 0.11 0.08 0.05 Availability of Electricity

(0.05) (0.07) (0.05) (0.05)

0.84 1.44** -0.06 0.83** Av. Education of Village Heads

(0.69) (0.59) (0.35) (0.36)

0.03* 0.01 0.00 0.01 Ethnic Diversity

(0.01) (0.02) (0.01) (0.01)

0.83 -0.69 -0.19 0.31 Natural Disaster Occurred

(0.65) (0.87) (0.65) (0.44)

-2.93*** -1.45* -0.61 -1.52***Tsunami: Physical Destruction

(0.58) (0.75) (0.48) (0.47)

-0.09 -0.08 -0.03 0.02 Average Neighboring Growth: Total GDP (0.08) (0.14) (0.06) (0.10)

-0.17 -1.63** -0.69** -0.85** Conflict: Material Damage (% GDP) (0.43) (0.76) (0.31) (0.35)

0.10 0.12 -1.15 -0.97 Conflict: Fatalities per Capita

(0.90) (1.57) (1.02) (0.84)

-0.70 0.99 -0.23 0.43 Spillover: Conflict in Neighboring Districts (0.62) (1.24) (0.98) (0.54)

0.29 1.48*** 0.96*** 1.06*** Time Dummy: Year 2005

(0.45) (0.33) (0.30) (0.21)

0.60 1.55*** 1.60*** 1.16*** Time Dummy: Year 2008

(0.41) (0.44) (0.30) (0.23)

2.61 -13.03* -2.13 -5.01 Constant

(6.11) (7.83) (5.71) (4.83)

Observations 1120 1123 1123 1123

R-squared 0.19 0.11 0.11 0.17

No. of Groups 375 375 375 375

Av. Obs. per Group 2.99 2.99 2.99 2.99

Robust standard errors in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%. Conflict Indices: Fatalities per 1,000 Inhabitants. Material Damage as 0.01 %-Points of Total District GDP.

34

Table 5: Regressions Results for Sub-sectors

(5) (6) (7) (8) (9) (10) (11) (12) Industry Services

Dependent Variable: GDP Growth

Mining Manufact. Energy Construct. Commerce Transport Finance Services 0.00 0.08 -0.02 0.14 0.01 0.01 0.04 -0.07

Population Growth Rate (0.15) (0.06) (0.15) (0.13) (0.08) (0.08) (0.12) (0.09)

0.17* 0.08 0.02 0.08 0.07 0.13** 0.29*** 0.01 Availability of Electricity

(0.10) (0.06) (0.12) (0.10) (0.06) (0.06) (0.09) (0.06)

-0.29 1.44*** 0.06 2.25*** 0.45 -0.47 -0.40 -0.24 Av. Education of Village Heads

(1.10) (0.42) (0.80) (0.75) (0.50) (0.55) (1.10) (0.58)

0.01 -0.02 -0.06* 0.04 0.01 0.02 -0.01 -0.02 Ethnic Diversity

(0.03) (0.02) (0.03) (0.02) (0.01) (0.02) (0.03) (0.02)

0.30 -1.81** 0.04 -0.38 -0.69 1.83 0.83 -1.50 Natural Disaster Occurred

(1.86) (0.83) (1.47) (1.17) (0.78) (1.14) (1.56) (1.01)

-0.98 -2.02*** -1.11*** -1.28 -1.05* -0.38 0.29 -0.54 Tsunami: Physical Destruction

(1.01) (0.32) (0.35) (0.94) (0.56) (0.72) (1.44) (0.62)

-0.23 0.02 0.03 -0.15 -0.07 -0.06 -0.04 0.05 Average Neighboring Growth: Total GDP (0.14) (0.14) (0.15) (0.14) (0.08) (0.07) (0.15) (0.07)

-0.26 -1.46*** 0.03 -1.22* -0.78** -1.44** -1.63** -0.37 Conflict: Material Damage (% GDP) (0.41) (0.51) (0.49) (0.71) (0.38) (0.66) (0.67) (0.29)

-0.81 1.35 -1.66 2.13 -1.08 -1.54* 0.88 0.62 Spillover: Conflict in Neighboring Districts (0.78) (1.40) (1.21) (2.08) (1.21) (0.89) (2.12) (1.39)

0.98 1.18*** -1.15* 1.49*** 1.58*** -0.19 0.12 0.56 Time Dummy: Year 2005

(0.72) (0.30) (0.60) (0.48) (0.38) (0.39) (0.79) (0.50)

0.98 1.69*** -1.14 1.86*** 1.77*** 0.67 1.85*** 1.73*** Time Dummy: Year 2008

(0.79) (0.44) (0.71) (0.61) (0.43) (0.48) (0.67) (0.50)

-10.50 -8.45 9.69 -14.22 -3.70 -5.67 -19.10* 7.12 Constant

(11.24) (6.52) (11.23) (10.40) (6.83) (7.29) (9.95) (6.94)

Observations 1036 1118 1115 1115 1121 1117 1104 1110

R-squared 0.04 0.13 0.02 0.07 0.09 0.10 0.08 0.04

No. of Groups 354 375 375 375 375 375 375 375

Av. Obs. per Group 2.93 2.98 2.97 2.97 2.99 2.98 2.94 2.96

Robust standard errors in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%. Conflict Index: Material Damage as 0.01 %-Points of Total District GDP.

35

Table 6: Regression Results – Resolved vs. Ongoing Conflicts

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Dep. Var: GDP Growth Agricult. Mining Manufact. Energy Construct. Industry Commerce Transport Finance Services Service Total

0.06 -0.00 0.08 -0.02 0.13 0.01 0.01 0.01 0.04 -0.07 -0.05 -0.08 Population Growth Rate (0.08) (0.15) (0.06) (0.15) (0.13) (0.09) (0.08) (0.08) (0.12) (0.09) (0.06) (0.07)

-0.05 0.17* 0.08 0.02 0.08 0.11 0.07 0.13** 0.29*** 0.01 0.08 0.05 Availability of Electricity (0.05) (0.10) (0.06) (0.12) (0.10) (0.07) (0.06) (0.06) (0.09) (0.06) (0.05) (0.05)

0.85 -0.30 1.44*** 0.06 2.21*** 1.42** 0.44 -0.49 -0.40 -0.23 -0.07 0.81** Av. Education of Village Heads (0.69) (1.10) (0.42) (0.81) (0.74) (0.59) (0.50) (0.55) (1.10) (0.58) (0.34) (0.36)

0.03* 0.01 -0.02 -0.06* 0.04 0.01 0.01 0.02 -0.01 -0.02 0.00 0.01 Ethnic Diversity

(0.01) (0.03) (0.02) (0.03) (0.02) (0.02) (0.01) (0.02) (0.03) (0.02) (0.01) (0.01)

0.83 0.24 -1.81** 0.03 -0.39 -0.69 -0.69 1.83 0.83 -1.50 -0.21 0.30 Natural Disaster Occurred (0.65) (1.87) (0.83) (1.47) (1.17) (0.87) (0.78) (1.14) (1.56) (1.01) (0.65) (0.44)

-2.94*** -0.98 -2.02*** -1.10*** -1.27 -1.44* -1.05* -0.37 0.29 -0.55 -0.60 -1.51***Tsunami: Physical Destruction (0.58) (1.00) (0.32) (0.34) (0.94) (0.75) (0.56) (0.72) (1.44) (0.62) (0.48) (0.47)

-0.09 -0.23 0.02 0.04 -0.09 -0.03 -0.06 -0.02 -0.04 0.03 -0.04 0.04 Average Neighboring Growth: Total GDP (0.08) (0.14) (0.15) (0.15) (0.15) (0.14) (0.09) (0.07) (0.17) (0.07) (0.06) (0.10)

-0.40 0.36 -1.36 0.17 1.06 0.04 -0.42 0.20 -1.44 -1.01*** -0.68* 0.12 Resolved Conflict: Material Damage (0.60) (0.69) (0.95) (1.05) (0.89) (0.62) (0.41) (0.33) (0.89) (0.34) (0.35) (0.37)

0.07 -1.75 -1.58** -0.13 -3.22*** -3.05*** -1.10 -2.92*** -1.79 0.19 -1.26** -2.16***Ongoing Conflict: Material Damage (0.46) (1.50) (0.65) (0.69) (0.78) (0.52) (0.68) (0.40) (1.44) (0.34) (0.59) (0.33)

-0.75 -0.55 1.38 -1.62 2.67 1.39 -1.00 -1.15 0.93 0.48 -0.23 0.65 Spillover: Conflict in Neighboring Districts (0.64) (0.93) (1.46) (1.26) (2.14) (1.29) (1.23) (0.92) (2.16) (1.35) (0.97) (0.55)

0.28 1.01 1.19*** -1.15* 1.58*** 1.55*** 1.60*** -0.12 0.13 0.53 0.98*** 1.12*** Time Dummy: Year 2005 (0.45) (0.73) (0.30) (0.60) (0.48) (0.33) (0.38) (0.39) (0.79) (0.49) (0.29) (0.21)

0.59 1.02 1.69*** -1.14 1.94*** 1.61*** 1.78*** 0.72 1.85*** 1.71*** 1.61*** 1.20*** Time Dummy: Year 2008 (0.41) (0.80) (0.44) (0.71) (0.60) (0.44) (0.42) (0.48) (0.67) (0.50) (0.29) (0.23)

2.67 -10.40 -8.46 9.65 -14.58 -13.35* -3.77 -5.97 -19.14* 7.25 -2.29 -5.34 Constant

(6.10) (11.23) (6.55) (11.20) (10.41) (7.80) (6.84) (7.31) (9.97) (6.93) (5.72) (4.82)

Observations 1120 1036 1118 1115 1115 1123 1121 1117 1104 1110 1123 1123

R-squared 0.19 0.04 0.13 0.02 0.08 0.11 0.09 0.10 0.08 0.04 0.11 0.17

No. of Groups 375 354 375 375 375 375 375 375 375 375 375 375

Av. Obs. per Group 2.99 2.93 2.98 2.97 2.97 2.99 2.99 2.98 2.94 2.96 2.99 2.99

Robust standard errors in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%. Conflict Index: Material Damage as 0.01 %-Points of Total District GDP.

36

Table 7: Robustness Tests

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Dep. Var: GDP Growth Agricult. Mining Manufact. Energy Construct. Industry Commerce Transport Finance Services Service Total

1SPATIAL SUB-SAMPLES a. Rural Areas

-0.41 0.51 -1.54 0.33 1.25 0.05 -0.42 0.46 -1.30 -1.29*** -0.70* -0.07 Resolved Conflict: Material Damage (0.50) (0.78) (1.04) (1.18) (0.84) (0.51) (0.46) (0.36) (0.81) (0.40) (0.40) (0.29)

-0.06 -1.84 -1.28** -0.22 -2.89*** -2.73*** -1.15 -3.20*** -1.86* 0.52 -1.25** -1.95***Ongoing Conflict: Material Damage (0.49) (1.76) (0.61) (0.82) (0.71) (0.48) (0.71) (0.40) (1.04) (0.49) (0.62) (0.32)

-0.92 -0.37 2.17 -2.43 3.61 1.78 -1.72 -2.16** -2.10* 1.66 -0.35 0.50 Spillover: Conflict in Neighboring Districts (0.85) (1.12) (1.83) (1.57) (2.92) (1.71) (1.79) (0.99) (1.12) (1.85) (1.45) (0.72)

b. Exclusion of NAD

0.43 0.96 -0.80 1.10 2.67* 0.56 -0.12 0.23 -2.33*** -0.31 -0.43 0.20 Resolved Conflict: Material Damage (0.90) (0.88) (1.04) (1.19) (1.54) (0.86) (0.55) (0.62) (0.89) (0.56) (0.44) (0.62)

-3.22 -4.03* -2.80 -5.24** -8.03** -4.45** -1.68 -3.15 2.34 -1.46 -1.36 -2.35 Ongoing Conflict: Material Damage (2.34) (2.30) (2.77) (2.63) (3.33) (2.14) (1.18) (2.16) (1.98) (1.57) (1.35) (1.63)

-0.59 -1.84 -0.53 0.42 -0.67 -1.03 -1.74* -1.42* -1.95* -1.34* -1.68** -0.59 Spillover: Conflict in Neighboring Districts (0.91) (1.28) (0.88) (1.22) (0.85) (0.82) (1.02) (0.75) (1.07) (0.73) (0.81) (0.62)

c. Exclusion of Sulawesi and the Moluccas

-0.06 0.47 -3.77*** 4.39** 4.39*** 0.05 -1.23 -1.22 -3.23 -1.84* -1.96** -0.56 Resolved Conflict: Material Damage (0.64) (1.58) (1.45) (2.18) (1.65) (1.03) (1.15) (0.99) (2.60) (1.05) (0.99) (0.59)

0.18 -3.10 -2.06** 0.43 -4.06*** -3.84*** -1.58** -3.14*** -3.08*** -0.27 -1.80*** -2.80***Ongoing Conflict: Material Damage (0.37) (2.94) (0.96) (0.77) (0.99) (0.70) (0.74) (0.36) (1.16) (0.35) (0.52) (0.33)

-1.00 1.19 2.93 -3.44 5.42 3.79* -0.04 -0.68 4.29 1.75 1.08 2.15*** Spillover: Conflict in Neighboring Districts (0.87) (1.67) (2.39) (2.29) (3.32) (1.96) (2.07) (1.45) (2.66) (2.47) (1.65) (0.73)

1ALTERNATIVE CONFLICT INDICATORS c. Fatalities per Capita

-1.88* -3.48 0.14 8.09 0.40 1.62 -1.28 0.90 1.26 -2.11 -0.96 0.30 Resolved Conflict: Fatalities (1.01) (2.31) (1.40) (7.36) (1.70) (1.24) (1.29) (1.20) (2.62) (1.31) (0.95) (0.95)

0.93 2.61 -2.97** -4.76 -4.20* -6.23*** -2.25 -4.91** -7.88*** -0.12 -3.17** -4.53***Ongoing Conflict: Fatalities (0.78) (3.04) (1.19) (4.51) (2.25) (1.55) (1.79) (2.33) (1.57) (1.43) (1.58) (1.35)

-0.63 0.40 1.82 2.06 3.84 3.13 -3.37* -2.66* 1.55 1.50 -0.82 1.10 Spillover: Conflict in Neighboring Districts (1.21) (1.70) (3.09) (6.07) (3.97) (2.47) (1.89) (1.38) (3.58) (2.40) (1.64) (1.27)

e. Share of Villages Affected by Conflict

0.57 0.69 0.30 0.86 0.58 0.54 -0.02 -0.24 -0.17 -0.17 -0.15 0.34 Resolved Conflict: Fatalities (0.47) (0.82) (0.51) (0.80) (0.63) (0.49) (0.35) (0.41) (0.61) (0.47) (0.26) (0.31)

0.31 -0.54 0.28 -1.55 0.61 -0.55 -1.48 -1.45 -2.30 0.11 -0.98 -0.51 Ongoing Conflict: Fatalities (0.35) (0.84) (1.31) (1.30) (2.12) (1.45) (0.94) (1.15) (1.62) (0.82) (0.87) (0.91)

0.08 0.72 0.38 -0.18 0.54 0.93** 0.12 0.16 0.29 -0.28 -0.05 0.49 Spillover: Conflict in Neighboring Districts (0.31) (0.57) (0.50) (0.56) (0.46) (0.47) (0.32) (0.42) (0.64) (0.47) (0.30) (0.31)

f. Material Damage – Absolute Values

-0.48 0.40 0.37 0.37 1.02 0.69 -0.29 -0.23 -1.64*** -1.18*** -0.66** 0.08 Resolved Conflict: Material Damage (0.48) (1.26) (1.06) (0.81) (1.04) (0.80) (0.32) (0.39) (0.54) (0.37) (0.25) (0.56)

0.19 -3.33 -0.65*** -0.04 -1.29*** -1.44*** -0.80*** -1.50*** -1.34*** 0.16 -0.80*** -1.09***Ongoing Conflict: Material Damage (0.13) (2.57) (0.22) (0.24) (0.22) (0.21) (0.13) (0.12) (0.33) (0.12) (0.11) (0.15)

-0.50 0.70 1.01 -1.00 1.80 1.15 -0.38 -0.50 0.99 0.29 -0.00 0.52 Spillover: Conflict in Neighboring Districts (0.33) (0.76) (0.95) (0.79) (1.36) (0.84) (0.74) (0.54) (1.20) (0.87) (0.58) (0.34)

g. Fatalities – Absolute Numbers

-0.37 -0.84** 0.31 2.61** 0.04 0.52*** -0.39** 0.26 0.46 -0.27 -0.21 0.06 Resolved Conflict: Fatalities (0.23) (0.40) (0.21) (1.08) (0.37) (0.20) (0.18) (0.16) (0.44) (0.25) (0.16) (0.16)

0.23 0.14 -0.86*** -1.36 -1.24* -1.95*** -1.00*** -1.86*** -2.37*** -0.14 -1.21*** -1.50***Ongoing Conflict: Fatalities (0.22) (1.31) (0.29) (1.21) (0.64) (0.38) (0.35) (0.43) (0.37) (0.39) (0.26) (0.29)

-0.24 0.30 0.23 1.23 0.73 0.67 -0.82** -0.61* 0.33 0.26 -0.28 0.16 Spillover: Conflict in Neighboring Districts (0.22) (0.33) (0.74) (1.52) (1.00) (0.59) (0.39) (0.32) (0.85) (0.50) (0.34) (0.33)

Each pair of coefficients from a different regression (control variables as in Table 5). Robust standard errors in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%. Conflict Indices: a) relative: Fatalities per 1,000 Inhabitants. Material Damage as 0.01 %-Points of Total District GDP. b) absolute: Number of Fatalities (in 100 deaths). Material Damage (in 1 billion Rupiah).

37

APPENDIX

Table I: Structure of the GDP Data

Sub-Sector Sub-Branches

GDP Share 2000 (Rural Areas)

GDP Share 2008(Rural Areas)

Farm Food Crops Estate Crops Livestock Forestry A

GR

ICU

L-

TU

RE

Agriculture

Fishery

16.7† (27.5)

15.5 (26.2)

Mining (Oil) Mining (Others) Mining Quarrying

11.1 (18.8)

8.5 (14.8)

Oil & Gas Manufacturing - Petroleum Refinery - Liquefied Natural Gas Non Oil & Gas Manufacturing - Food, Beverage & Tobacco - Textile, Leather Products & Footwear - Wood Products & Forest Products - Paper - Fertilizer, Chemicals & Rubber Products - Cements and Non-metal Quarrying Products - Iron & Steel - Transport. Vehicle, Machinery & Equipment

Manufacturing

- Other Manufactured Products

26.3 (21.7)

24.3 (22.2)

Electricity Gas Energy Water Supply

1.0 (0.9)

1.1 (1.0)

IND

US

TR

Y

Construction Construction Total 5.1

(3.3) 5.9

(3.9) Large Trade & Retail Hotel Commerce Restaurant

17.5 (14.5)

20.0 (16.5)

Transportation - Railway Transport - Road Transport - Sea Water Transport - Inland Water Transport - Air Transport - Transportation Supporting Communication - Post and Telecommunication

Transport & Communication

- Communication Supporting

5.2 (3.4)

6.9 (4.0)

Banking Non-banking Financial Institution Financial Supporting Rent of Building

Finance

Business Services

8.6 (2.7)

8.8 (3.1)

General Government - Government Administration & Defense - Other Government Services Private - Social Community - Recreation & Entertainment

SE

RV

ICE

S

Services

- Household & Private

8.5 (7.2)

9.0 (8.2)

† Percentage share of total (rural) GDP.

38

Table II: Unchanged GDP Data

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Dep. Var: GDP Growth Agricult. Mining Manufact. Energy Construct. Industry Commerce Transport Finance Services Service Total

a. All Areas

-0.74 -6.92 -1.54 0.02 0.27 -1.25 -0.35 0.10 0.20 -2.66 -1.11 0.10 Resolved Conflict: Material Damage (0.57) (5.76) (1.00) (1.14) (1.06) (1.65) (0.42) (0.42) (1.86) (3.52) (1.36) (0.68)

0.23 7.82 -1.60** 0.21 -2.10** -2.07 -1.09 -2.45*** -0.23 -0.96 -1.62** -3.79***Ongoing Conflict: Material Damage (0.55) (5.48) (0.67) (0.83) (0.90) (2.05) (0.66) (0.58) (3.52) (2.80) (0.79) (1.19)

-0.91 -16.04 1.28 -2.11 -0.04 0.27 -1.05 -0.18 -10.51 -4.16 -1.05 2.47 Spillover: Conflict in Neighboring Districts (0.90) (16.33) (1.46) (1.78) (2.52) (4.85) (1.21) (0.94) (11.07) (5.11) (1.76) (3.28)

b. Rural Areas

-0.90* -7.07 -1.81* 0.06 0.50 -1.61 -0.37 0.20 2.82 -2.48 -1.02 -0.33 Resolved Conflict: Material Damage (0.49) (5.82) (1.10) (1.22) (1.04) (2.10) (0.47) (0.47) (2.76) (3.75) (1.43) (0.82)

0.10 7.95 -1.29* 0.29 -1.99** -1.35 -1.13 -2.55*** -3.00 -1.30 -1.73** -3.31***Ongoing Conflict: Material Damage (0.58) (5.19) (0.67) (0.95) (0.90) (2.44) (0.70) (0.58) (2.21) (3.06) (0.88) (1.16)

-1.32 -19.51 2.00 -3.38 0.13 1.07 -1.77 -0.38 -18.10 -6.99 -1.99 3.10 Spillover: Conflict in Neighboring Districts (1.16) (22.69) (1.85) (2.38) (3.54) (6.88) (1.77) (1.34) (14.31) (7.19) (2.57) (4.56)

c. Fatalities per Capita

-1.33 -1.21 -0.30 8.23 1.23 2.82* -1.00 1.68 3.35 -4.06 -1.60 1.68 Resolved Conflict: Fatalities (1.19) (2.39) (1.61) (7.24) (1.49) (1.68) (1.29) (1.30) (3.50) (4.62) (1.53) (1.52)

1.30 14.10** -2.09 -2.96 -3.67* -2.05 -3.08** -4.78** -11.19* -0.80 -3.40* -6.32***Ongoing Conflict: Fatalities (0.94) (6.40) (1.63) (4.29) (2.18) (2.40) (1.53) (2.23) (5.98) (6.85) (1.91) (1.62)

-1.02 -11.87 0.80 -0.14 0.60 2.10 -3.30* -1.13 -15.21 -28.39* -7.55* 2.13 Spillover: Conflict in Neighboring Districts (1.58) (16.32) (3.04) (7.29) (4.23) (7.98) (1.96) (1.35) (19.19) (16.72) (4.23) (5.87)

Each pair of coefficients from a different regression (control variables as in Table 5). Robust standard errors in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%. Conflict Indices: Fatalities per 1,000 Inhabitants. Material Damage as 0.01 %-Points of Total District GDP.